Site-controlled quantum dot growth on buried oxide stressor layers



Site-controlled growth of quantum dots (QDs) for single photon emitters (SPEs) is achieved applying a buried stressor approach. Theoretical and experimental analysis shows that site-controlled QD growth on buried oxide stressor-layers benefits enormously from a defect-free growth interface. Laterally modulated strain fields at GaAs(001) growth surfaces are used to tailor surface morphologies at the centre of prescribed mesa structures for subsequent QD growth. Suitable morphologies for site-controlled QD growth such as nano-hillocks and nano-holes are identified. Site-controlled QD growth appears above the boundaries between the oxidised layer and the non-oxidised semiconductor layer. Through fine tuning of wetting layer thickness and growth interruption high selectivity for QD nucleation is achieved. Thus, growth of single QDs at the centre of a current-injection limiting aperture is demonstrated. Moreover, the QD growth on a defect-free surface yields high quality optical properties in terms of narrow emission linewidth and temporal stability with no discernible difference to QDs grown on planar substrates. The technological simplicity of the buried stressor approach and the inherent integration of a current aperture for efficient carrier injection into site-selected QDs enable mass production of SPEs on large substrate sizes.

1 Introduction

Single-photon emitters (SPEs) are key elements of future secure data communication based on quantum cryptography 1. Electrically driven miniaturised SPEs based on semiconductors are particularly attractive for mass production and are already demonstrated using single QDs as single-photon source 2. Quantum cryptography as well as quantum computing requires manipulation of quantum states (so-called QuBits) for which long coherence length of photons is of fundamental importance. Sources of emission-line broadening which reduce the coherence time need therefore to be avoided. Highest optical quality of quantum dots (QDs) is achieved by QD growth on planar substrates in the Stranski–Krastanow regime 3, 4. A homogenous linewidth of In0.6Ga0.4As QDs as small as 2 µeV at low temperatures was found under resonant optical excitation 5. Until recently, electrically driven SPEs are demonstrated using randomly distributed QDs grown in the Stranski–Krastanow regime on planar surfaces 6–11. While excellent performance is achieved regarding single photon generation, the device yield per wafer is limited since current injection into a single QD remains a matter of luck as there is no correlation between the position of a QD and the current path defined by device geometries. Therefore, much research is directed towards site-controlled QD growth where the QD nucleation is determined by local surface structures. Unfortunately, planar surfaces of conventional semiconductor substrates do not exhibit features suitable for site-control of single QD nucleation. Site-controlled growth of QDs with nano-scale precision is only possible if the surface is altered on a nanometer scale. Thus, various nano-lithographic technological approaches to define QD nucleation sites were previously demonstrated, e.g. 12–19. Common to these top-down approaches is direct surface patterning by means of masking, etching, or indentation leading to holes, inverted pyramids, or simply to selective growth windows. They also share a number of difficulties, which impact the structural and optical properties of QDs. One problem is the very close vertical proximity to the structural patterning required for deterministic QD nucleation because of the short-range impact of the nano-structured surface features. Missing of QD nucleation at shallow holes patterned on a growth surface is often reported 15, 16. Another one relates to the surface damage by etching or other surface invasive means whereby the QDs will be surrounded by defect sites which degrade their structural and optical quality 5. Some progress to solve these issues was made using sophisticated cleaning procedures prior to growth and QD stacking. Still, the majority of the site-controlled QDs show inferior optical qualities in terms of spectral linewidth and radiative efficiency as compared to QDs grown on planar substrate surfaces 15–17. Minimum spectral QD linewidths of 43, 80 and 100 µeV are occasionally demonstrated for individual site-controlled QDs 15, 17, 20. Recently, electrically driven SPEs based on site-controlled QD growth are demonstrated. The linewidth of the excitonic emission of 650 and 170 µeV indicate persistent broadening due to the pre-patterning process 21, 22.

Another major drawback of these technologies, however, is their limited scalability to large wafer sizes because of the involved delicate nanometer-scale processing steps typically requiring sequential rather than parallel processing which is at least very time-consuming. Furthermore, for efficient electrical injection, precise alignment of the electrical current path with respect to the QDs remains a not so easy technological effort as most of the nano-lithographic techniques provide no self-alignment of the current injection path to site-controlled QDs.

Solutions to the current problems of site-controlled QD growth require both long-range vertical ordering effects as well as an easily scalable surface preparation technology. Aiming both at large scale processes and defect-free growth surfaces, the buried stressor approach was recently developed by us 23. Long range, site-controlled QD growth is achieved through laterally modulated strain fields at the surface of a GaAs(001) substrate. The novel technique does not require nano-lithographic surface patterning and avoids all problems associated with etching, indentation or masking. Instead, laterally oxidised AlAs layers are used to create a locally varying strain field at the centre of >10 µm large mesa structures. Sub-micron precision comparable to the nano-lithographic approaches only requires control of the depth of the lateral oxidation. Inherent full wafer scalability as well as self-aligned vertical current path renders the approach very attractive to mass production on large wafer sizes.

In this paper, details on modelling, preparation of different surface morphologies, selectivity of QD nucleation and spectral analysis of optical properties are reported. Details of the buried stressor formation through selective oxidation of AlAs layers are given first. The most important effect of the oxidation process is the laterally modulated strain field at the GaAs (001) growth surface which, as shown in Section 2, is accurately modelled by continuum mechanical treatment. Section 3 follows with results of growth experiments which demonstrate different attainable surface morphologies prior to QD deposition, possibilities of QD alignment and selectivity of QD nucleation, respectively. Optical properties of individual QDs supporting the defect-free growth of QD will be presented in Section 4 before the prospects of the buried stressor approach are discussed.

2 Properties of the buried oxide layer

2.1 Oxide formation

Oxide formation through lateral oxidation of buried AlAs layers is known since 1978 and has been used as electrically and/or optically confining elements in vertical cavity surface emitting lasers (VCSEL), and as low index material in broadband distributed Bragg reflectors 24–26. The oxidising agent is water vapour which leads to chemical reactions replacing the anion arsenic with oxygen. Choquette et al. describe the reaction as follows 27:

equation image((1))

This is an exothermic reaction with a change of the Gibbs free energy of −473 kJ mol−1. For oxidation of GaAs, however, the reaction is endothermic which leads to high selectivity of the oxidation between GaAs and AlAs layers. The oxide formation is known to proceed without defect generation in surrounding semiconductor layers. Experimentally, one finds a linear reduction of the layer thickness by about 12–13% upon oxidation in AlAs layers depending on Al content and process conditions 28. For Al0.92Ga0.08As layers, the thickness reduction is only 6.7% 29. Strain-fields are identified by transmission electron microscopy for AlAs/oxide interfaces while strain-free interfaces are observed for Al0.98Ga0.02As/oxide interfaces 30. Stressors for QD positioning should therefore contain AlAs as oxidation layers.

The epitaxial structure for the buried stressor comprises a sandwich of Al0.9Ga0.1As/AlAs/A0.9Ga0.1As layers which is buried beneath 50 nm of GaAs. Circular mesas of 300 nm height and with stepwise decreasing diameters from 25 to 10 µm are formed by photolithography and dry etching. The oxidation process is initiated at 425 °C using 50 mL water vapour in 3 slpm nitrogen carrier gas. In situ monitoring of the oxidation front is accomplished by surface inspection using a camera attached to a microscope. Thus, stopping of the oxidation front with a precision of about 1 µm is possible. Sub-micron apertures are obtained by stepwise decreasing mesa diameters as implemented in the mask layout. Figure 1a shows the cross-section scanning electron microscopy (SEM) image of the stressor layer structure after oxidation. The oxidation proceeds within the AlAs layer and a well-defined rectangular boundary is formed to the non-oxidised AlAs. The GaAs surface morphology after oxidation as recorded by atomic force microscopy (AFM) can be seen in Fig. 1b. The contour of the oxidation front is well resolved as a surface undulation with an amplitude of about 1–2 nm. For large apertures, square-like apertures with rounded corners are obtained due to the anisotropy of the oxidation process. When the aperture size approaches sub-micron dimensions the contour becomes circular as the rounded corners merge.

Figure 1.

(online colour at: (a) Cross-section SEM image of the stressor layer showing the oxide/AlAs interface and the aperture where the AlAs/AlGaAs sandwich is un-oxidised. The inset shows the circular mesa structure after oxidation. (b) Surface of the centre of a mesa as seen by AFM. The dashed line denotes the square-like aperture in the oxide.

2.2 Strain distribution

The strain distribution across the surface is modelled by continuum elasticity theory in a finite-differences framework. Using a conjugated-gradient approximation the energy of the solid with respect to strain is minimised. The internal forces arise from contraction of the stressor layer volume during oxidation. These forces pull the aperture towards the mesa edge. Thereby, tensile strain is created within the aperture region.

The growth surface can be characterised by the sum equation image of the local surface strains. This quantity determines the strain energy which builds up in subsequently grown layers. In the following, tuning parameters for the surface strain profile are obtained by analysing structural parameters such as the aperture size, the mesa size, the vertical distance of the surface to the stressor layer and the shape of the aperture boundary. Important aspects of the stressor design for site-controlled growth of QDs are the dependence of the strain distribution on the aperture size and the magnitude of the strain at the surface. For this purpose, Fig. 2 shows strain profiles taken across a circular mesa of 3 µm diameter with apertures of 0.5–1 µm size. The magnitude of the local strain can reach 1–2% which is of the same order as the lattice mismatch strain of InGaAs/GaAs(001). Therefore, the energy balance for Stranski–Krastanow growth of QDs is locally modified due to the surface strains and local QD growth can be expected. The largest modulation of the surface strains and therefore largest impact on the energy balance is close to the aperture boundaries.

Figure 2.

(online colour at: Line scans of surface strains equation image across a circular mesa of 3 µm diameter exhibiting different aperture sizes. The horizontal line marks unstrained GaAs.

For larger apertures (>1 µm), the profile exhibits an outer surface region of compressive strain surrounding the aperture. At the aperture boundaries on both sides the strain changes its sign from compressive to tensile reaching a maximum of tensile strain at the inner side of each aperture boundary. Across the aperture the strain remains tensile but of lower magnitude. Now, with shrinking aperture diameter, the two maxima of tensile strain approach each other (see Fig. 2) and finally merge into one tensile strain maximum at the centre of the aperture for sub-micron aperture diameters. This always coincides with the centre of the mesa independent of the mesa diameter. Since the width of the tensile strain maximum can be as small as 300 nm single QDs may be grown with nm-precision at the mesa centre.

Figure 3a shows strain profiles for structures with the same aperture size of 0.5 µm but decreasing mesa sizes from 12 to 2 µm. The single tensile strain maximum always stays at the centre of the mesa as expected. The impact of the mesa size on the strain characteristics of the aperture boundary is rather weak for mesa sizes larger than 2 µm. The magnitude of the tensile strain is depending on the mesa size if it is decreased below 4 µm. Outside of the aperture, the surface strain is always compressive close to the aperture boundaries. Towards the mesa edge, it tends to approach zero strain between the mesa edge and the aperture boundary. At the mesa edge again a transition from compressive strain to tensile strain takes place. These two strained regions (e.g. the one at the mesa edge and the one at the aperture boundaries) overlap for mesa sizes below 4 µm resulting in two distinct domains of different strain.

Figure 3.

(online colour at: Simulation of the surface strain distribution equation image assuming a 7% layer thickness reduction. (a) Impact of the mesa size on the strain distribution. The tensile strain fields across the aperture region are affected by the compressive strain fields from the mesa edges for mesa sizes smaller than 4 µm. (b) Impact of surface to stressor distance. Beyond 100 nm vertical separation, the tensile strain field becomes a single maximum at the aperture centre. Even for 300 nm separation, significant strain is present at the surface.

In Fig. 3b the vertical distance of the QD growth surface to the stressor layer is increased from 30 to 300 nm. The corresponding mesa and aperture sizes are 3 µm and 500 nm, respectively. Below 100 nm GaAs cap layer thickness, local maxima of tensile strain can be resolved. For larger thicknesses, only a single maximum is located at the aperture centre, the transition from the compressive to tensile strain becomes less steep and the amplitude of the strain modulation is reduced. However, even at a distance of 300 nm from the stressor layer a significant tensile surface strain is maintained.

2.3 Growth model

QD formation due to strained layer growth in the Stranski–Krastanow mode is a consequence of total energy minimisation 3. The critical layer thickness for the 2D/3D growth mode transition depends on the interfacial lattice mismatch which depends on the composition of the growing layer. Ripening of QDs occurs because of material transfer from the wetting layer and small QDs. Site-controlled QD nucleation within the InGaAs/GaAs(001) material system in the Stranski–Krastanow regime may therefore be the result of a thickness and/or compositional variation within the wetting layer, and/or of a directed material transfer towards energetically favourable sites during QD formation. As the modulated strain field contributes to the surface free energy the growth of subsequent layers may vary locally in two ways. First, the growth rate may depend on the local surface position since atoms will migrate towards the local minimum of strain energy. Ternary compound layers may additionally undergo phase separation to accommodate for different strain states of the surface. For example, InGaAs layers on a strain-modulated GaAs surface are likely to show higher than the nominal In content in areas of tensile strain and lower In content in compressively strained GaAs regions. Both effects would contribute to a locally dependent 2D/3D transition layer thickness. High selectivity of the QD nucleation on the strained surface will require a wetting layer thickness close to the onset of the 2D/3D transition in order to reinforce the effects of the layer variations. The selectivity may also have a compositional dependence in the InGaAs/GaAs system as higher strain in the wetting layer may increase phase separation tendencies during growth.

The kinetics of QD formation in the InGaAs material system is driven by In adatom migration on the surface during the QD ripening process. During the ripening process In adatoms will also preferentially migrate towards tensile strained areas where they can minimise their free energy. The difference to a locally varying critical layer thickness is that the selectivity of the QD nucleation becomes time dependent during the ripening process. Figure 4 illustrates the situation for growth of GaAs (left panel) and InAs (right panel) layers on top of a strain-modulated GaAs(001) surface. Areas of low strain of the growing material are represented by dark colours. The areas of lowest strain for InAs layers are enclosed by red lines for enhanced visibility. Thus, it becomes obvious that InAs material will agglomerate preferably at the aperture boundaries and in particular at the corners of square-like apertures. The preferred GaAs growth region is adjacent but outside the aperture boundaries.

Figure 4.

(online colour at: Strain within a growing layer calculated with respect to GaAs (left) and InAs (right) deposited on top of a buried stressor layer. Dark areas mark the points of lowest strain energy. The areas enclosed by the red lines in the right panel show the regions of lowest strain for InAs.

3 Growth experiments

3.1 Preparation and quantum dot growth

The de-oxidation of GaAs(001) substrates prior to MOVPE growth is usually achieved in situ around 700 °C in arsine atmosphere. In order to maintain thermal stability of the oxidised mesa structures the oxide layer thickness is kept below 50 nm and the mesa diameters below 30 µm. The processing steps applied to the mesa surface prior to QD growth involve only photo-resist coating/removal, and rinsing in de-ionised water. This allows for handling of the mesa surfaces similar to epi-ready wafer surfaces. After de-oxidation, mono-atomic steps are observed on the surface, only. Due to the long-range impact of the stressor strain field on the GaAs surface it is possible to grow GaAs buffer layers of up to 150 nm thickness before QD deposition. Thus, the QD plane is very well separated from the interface of the stressor structure where non-radiative recombination centres might be present. For precise calibration of the transition layer thickness, the QD density on planar GaAs(001) substrates is reduced to below equation image. Since the effective area of the mesa-structured GaAs wafer surface for wetting layer deposition is effectively unchanged as compared to planar wafers the deviation of the QD density on mesa surfaces to planar wafer surfaces is only a factor 3–5 which means a thickness control of better than 0.1 ML.

All epitaxial growth runs are done in an commercial horizontal AIX-200/4 reactor. Growth rates and compositions of the binaries InAs, GaAs, AlAs are calibrated by measuring X-ray diffraction spectra of AlAs/GaAs and InGaAs/GaAs short period superlattices. Individual parameters for growth of certain layers are interpolated from these calibrations. After de-oxidation, the growth of GaAs buffer layers is performed at 700 °C using standard metalorganic group-III precursors and AsH3 as group-V precursor. Afterwards, the reactor temperature is lowered to 515 °C for QD deposition. InGaAs is grown at a V/III ratio of 1.5 using tertiarybutylarsine as group-V precursor. Subsequently a growth interruption without group-V supply is applied. For AFM, the QDs are capped by a 6 nm thin GaAs cap layer to freeze the QD alignment. For optical characterisation, the sample structure comprises additional 50 nm GaAs and a 20 nm/5 nm AlGaAs/GaAs charge carrier diffusion barrier on top of the QDs.

3.2 Surface morphologies

As pointed out in Section 2.3 the stressor layer may lead to a growth rate variation across the mesa surface. Figure 5 shows GaAs surfaces of mesas with different diameters after 100 nm GaAs growth at 700 °C which corresponds to the surfaces right before QD deposition. Mono-atomic surface steps are clearly resolved demonstrating perfect two-dimensional growth behaviour of the GaAs buffer layer. The density as well as the curvature of the mono-atomic steps refers to the GaAs growth rate. Far away from the aperture, the average step distance is about 400 nm as seen in Fig. 5a. At the aperture boundaries, the average step distance reduces to less than 200 nm indicating at least a factor of two higher GaAs growth rate. Surrounding the aperture is a rectangular area where hardly any step is observed which means a very low growth rate. We attribute this area to the region of maximum compressive strain.

Figure 5.

(online colour at: Surface morphologies of mesas with different aperture sizes after growth of 100 nm GaAs. The step morphology is indicative of growth rate variations. (a) Square-like aperture, (b) sub-micron hillock aperture, (c) sub-micron hole.

For apertures >1 µm the aperture shape is rectangular with rounded corners. Sub-micron apertures lead to more or less circular hillocks of 3–5 nm height (Fig. 5b). For very small apertures <0.2 µm (Fig. 5c), GaAs growth results in a well-defined hole in the surface. The observed transition from square-like to circular aperture shapes for aperture sizes <1 µm and the GaAs growth rate variations are in good agreement with the modelled strain distributions and the anticipated impact on layer growth.

3.3 Quantum dot site-control

In the following, the surface regions on the mesa are denoted as mesa edge, oxide region and aperture region according to the relative position above the stressor layer. On each mesa QD alignment is found at the mesa edge and at the aperture boundaries in agreement with the growth model derived from the strain distribution. The alignment region of QDs at the mesa edge is always extended along the edge and of little interest for the purpose of SPEs. In Fig. 6 AFM images of different aperture regions with decreasing sizes are shown. The aperture in Fig. 6a is square-like with a side length of 1 µm. A sub-micron circular aperture is seen in Fig. 6b whereas the aperture in Fig. 6c has the shape of a hillock of only a few hundred nm. The basics of the impact of the stressor layer become obvious in Fig. 6a. QDs preferentially nucleate at the square-like boundaries of large apertures. There, the areal density of QDs is about 1010 cm−2 while outside the QD density is only equation image, i.e. similar to the QD density as calibrated on planar substrates. For sub-micron circular apertures as in Fig. 6b, clusters of QDs nucleate in the centre of the mesa. Only two QDs nucleate on the smallest hillock-like apertures seen in Fig. 6c.

Figure 6.

(online colour at: QD alignment for different aperture sizes and shapes; (a) square-like arrangement, (b) QD cluster on hillock-like aperture, (c) single QDs.

The QD distribution depends on the average QD density. Figure 7 shows QD distributions across 15 µm large mesas for average QD densities of mid-109 cm−2 (Fig. 7a) and low-108 cm−2 (Fig. 7b). The QDs are formed after depositing In0.65Ga0.35As and a growth interruption of 5 min. The nominal difference in layer thickness is about 0.1 ML. Both samples exhibit a square-like region of enhanced QD nucleation above the aperture boundaries similar to the QD alignment seen in Fig. 6a. However, the selectivity of QD nucleation clearly depends on the average QD density. The low-density sample (Fig. 7b) exhibits only a few QDs outside the aperture region whereas for the high-density sample (Fig. 7a) there is only a narrow QD-free zone separating the aperture region from the oxide region. In general, one finds QD densities on the oxide region similar to planar surfaces in good agreement with the strain model predicting zero surface strain in the intermediate region between the aperture boundaries and the mesa edge.

Figure 7.

(online colour at: AFM images (deflection mode) of QD distributions across 15 µm large mesas for different QD densities. (a) mid-109 cm−2 and (b) low-108 cm−2. The QD-free zone around the aperture region is always present.

Ripening of QDs occurs during growth interruptions applied after initial wetting layer deposition. In general, material is transferred from smaller QDs to larger QDs. Growth interruptions can be used to tailor QD properties in particular their emission energies. Typical ripening times are of the order of a few to a few hundreds of seconds. As seen in Fig. 8, QDs nucleate first at the aperture boundaries after 30 s of growth interruption. After 60 s, a second set of QDs nucleate in an area surrounding the boundaries but separated from the initial QDs by a QD-free area. After long growth interruption of 120 s, QDs are found everywhere except in this QD-free area, proving that the critical layer thickness is reached both in regions of tensile and zero surface strain for a QD density of 108 cm−2. In order to achieve exclusively site-controlled QD nucleation the nominal QD density must be below 108 cm−2.

Figure 8.

(online colour at: Evolution of the QD distribution during growth interruptions of (a) 5 s; (b) 30 s; (c) 60 s; (d) 120 s. AFM height profiles.

By increasing the indium content of the InxGa1 − xAs QDs from around 40 to 100% the selectivity might increase since pure InAs may grow preferentially at the sites of highest tensile strain. Experimentally, no significant difference in selectivity of QD nucleation for samples with 45, 66 and 100% indium content is observed. It can be assumed therefore, that the selectivity of the QD nucleation results from the InAs material transfer during the growth interruption.

Optimum selectivity of QD nucleation is achieved for In0.66Ga0.34As layers and 15 s growth interruption albeit this might not be an exclusive set of parameters. Fine tuning of the layer thickness during consecutive runs on identical mesa structures is applied in order to maximise the selectivity of QD nucleation. The result is shown in Fig. 9, where exclusive QD formation at the points of highest tensile surface strain, i.e. at corners and sub-micron hillocks is observed. Due to the ultra-high selectivity and the inherently precise alignment with respect to a vertical current path in an oxide-aperture-confined pn-junction, device-efficient injection of charge carriers into the QDs can be expected.

Figure 9.

(online colour at: QD positioning with ultra-high selectivity. (a) Single QDs nucleate at the corners of the square-like apertures, only. (b) Single QDs at the centre of sub-micron hillock-like apertures. AFM images taken in deflection mode.

4 Optical properties

4.1 Ensemble characterisation

Low-temperature (T = 5 K) cathodoluminescence spectroscopy is employed to analyse spectral properties of QD related luminescence across the mesa. Circular mesas with diameters increasing from 10 to 25 µm (Δdmesa = 0.07 µm) are patterned on the sample. QDs are formed after growth of 50 nm GaAs followed by deposition of 3 ML of In0.66Ga0.34As and by applying a growth interruption of 60 s. Without QD formation the InGaAs layer forms a quantum well (QW) within the GaAs matrix. Two reference structures grown on a planar substrate as well as on a reference mesa template yield QD densities of around equation image as found by AFM measurement. From the reference mesa template the critical diameter for the onset of the stressor effect on the QD nucleation is determined to 15.5 ± 0.5 µm. As indicated in Fig. 10, QD luminescence is distributed in an energy interval between 910 and 1040 nm (1.34–1.28 eV) while InGaAs QW layer luminescence is observed at 895 nm (1.384 eV) (see integral spectrum). Different QD ensembles can be distinguished by their spectral footprint (QDI, QDII, QDIII). The ensembles QDI and QDII are mainly found at the aperture region while ensemble luminescence QDIII is found both at the mesa edge and at the aperture region. According to the spectral ordering from shorter to longer wavelengths the size of the QDs within the different ensembles is increased. The CL intensity maps shown in Fig. 10 are restricted to the QD emission bands in order to reveal the spatial distribution of QD luminescence. Across each mesa intense QD luminescence is found at the mesa edge. On top of the oxide region the QD luminescence intensity is low because of the lower QD density. At the mesa centre, the intensity of the QD luminescence depends on the QD site-control due to the stressor layer. Small mesas (<15 µm) do not have an aperture in the oxide and exhibit no distinct surface strain at the centre. Consequently, the QD distribution is homogeneous, has a low density and the QD luminescence intensity is weak. At around 15 µm mesa diameter, an aperture in the oxide is present and QD nucleation at the mesa centre is preferred. Simultaneously, the QD luminescence increases and the QW peak intensity is reduced. For larger mesas, the intense QD luminescence peak shows a square-like spatial distribution which compares well with the square-like arrangement of QDs seen by AFM (cf. Fig. 6a–c). At low QD densities, material transfer during QD formation from the QW to QDs is not expected to alter the emission energy of the QW. Thus, the energetic position of the QW luminescence can be assigned to thickness and composition of the QW as a result of the growth itself. Local variations of these parameters due to strain-modulated local growth rates may be deduced by monitoring the QW emission energy for different stressor sizes. For this purpose, CL linescans across mesas with increasing diameter from 14 to 16 µm as shown in Fig. 11 are recorded. Only in Fig. 12 minor shifts in QW emission energy of the order of <10 meV are observed indicating a weak dependence of the QW thickness and composition on the local surface strain. For 14 µm mesa diameter no strain modulation across the mesa centre is present due to complete oxidation of the stressor layer. At 15 µm diameter, the onset of the stressor effect is observed. Eventually at 16 µm diameter, QDs are nucleated at the mesa centre. During this transition the strain at the centre of the mesa changes from more or less unstrained to tensile. A significant change of several tens of meV in the spectral position of the QW peak is to be expected, if there is a change in the QW thickness due to growth rate variation. In contrast, the spectral position of the QW luminescence shifts only from 1.389 eV (14 µm mesa) to 1.381 eV (15 µm). From this observation material transfer during growth interruption can be concluded as the dominant process for selective QD nucleation.

Figure 10.

(online colour at: Cathodoluminescence intensity maps of the QD luminescence of circular mesas with increasing aperture size. (a) Aperture closed, no QD-related intensity at centre; (b) sub-micron aperture, cluster-related QD luminescence at centre; (c) square-like aperture, QD luminescence intensity highest on aperture boundary. The outer ring of luminescence intensity is attributed to QDs located at the mesa edge. The lower panel shows integral spectra taken at the mesa edge and at the aperture boundary as indicated by the dots in the schematic. QDI, QDII, and QDIII denote luminescence from different QD ensembles, whereas QW denotes luminescence of the InGaAs QW.

Figure 11.

(online colour at: CL linescans across mesas with increasing mesa diameter. The QW emission energy shows only little shift due to the stressor formation indicating homogeneous QW thickness and composition.

Figure 12.

(online colour at: CL transients of QD emission lines. Temporal jitter and blinking is not detected.

4.2 Single dot spectroscopy

Linewidth broadening due to spectral diffusion is one of the limiting factors for site-controlled QDs to be useful for quantum cryptography and quantum computing. Since the QDs are separated from the buried stressor interface by a 100 nm thick GaAs layer emission linewidths comparable to QDs grown on planar substrates are expected. Temporal analysis of single emission lines is performed by CL spectroscopy in order to detect spectral diffusion and possible underlying mechanisms. Figure 13 shows a series of 500 QD spectra emissions each integrated over 100 ms. The spectra are taken at the aperture region where selective QD nucleation occurs. Emission linewidths of individual lines of about 140 µeV are found which corresponds to the resolution limit of the CL setup. Most importantly, the luminescence of site-controlled QDs at the aperture region is free of noticeable temporal jitter or blinking. This result demonstrates the high optical quality of the QDs and the high structural quality of the GaAs matrix material, both spectral properties are known as sources of linewidth broadening.

Figure 13.

(online colour at: Micro-PL spectrum of a site-controlled QD emission line above the aperture. The inset shows the photon correlation measurement demonstrating anti-bunching.

Using a high resolution micro-photoluminescence setup QDs with linewidths below 100 µeV are identified at the aperture region. Figure 13 shows an example of a QD emission line with 65 µeV full width at half maximum (FWHM) on which anti-bunching experiments are performed using a Hanbury–Brown–Twiss setup. The photon-correlation measurement yields a normalised two photon coincidence of 0.35 at zero time delay. By deconvoluting the time response of the photodetectors (0.6 ns) a g(2)(0) = 0 is calculated which proves single photon emission from a single QD. Minimum FWHM of 40 µeV (resolution limit of the µ-PL setup) for some QD emission lines are found. Such values are comparable to InGaAs QDs grown by MOVPE on planar GaAs(001) substrates 31. Further improvement can be expected as many growth parameters are yet unexplored to improve the emission linewidth.

5 Summary

Buried stressor layers provide practical solutions to many fundamental technological and physical problems associated with site-controlled QD growth. Foremost, large-scale, high-yield fabrication of SPEs becomes possible since stressor preparation requires only µm-resolution lithographic tools and self-aligned vertical current injection is conceptually inherent. Site-controlled single QD growth is achieved by sub-micron aperture sizes along with properly defined QD growth parameters such as wetting layer thickness and growth interruption. Second, the ability to grow QDs on high quality thick buffer layers resolves issues of defect-related linewidth broadening of the QD emission. In principle, resulting optical properties of the QDs are only affected by the growth itself. Single photon emission and QD emission linewidths down to 40 µeV are demonstrated with potential for further improvement. Thus, the buried stressor approach has a huge potential for SPE devices. The seamless integration of a buried stressor layer into resonant-cavity SPE structures providing simultaneously current-confinement to the QDs due to the insulating properties of the oxide is essential. Such buried stressor layers may also be applied in nitride based and phosphide based material systems where AlInN and AlInAs layers can be selectively oxidised in water vapour 32, 33.


Funding by DFG within the collaborative research centre Sfb787 “Nano-Photonic–Materials, Models, Devices” is acknowledged.